My Thesis Behind Rogo

Rogo is building Wall Street's first "AI Analyst" to streamline current analyst workflows.

For many people, myself included, finance is seen as an exciting career where you can advise companies on high-stakes transactions and tackle some of the most pressing and current business problems of our time! As a second-year investment banking analyst – I would wholeheartedly agree with this view. In just over a year, I’ve been involved in several marquee transactions including the $8bn restructuring of an Apollo portfolio company, the $20bn sale and subsequent activism defense of Frontier Communications and worked with the NFL on developing a framework to allow private equity investments in teams.

Even before I started my investment banking role, I knew the inherent value of being a "fly on the wall" during high-level business meetings. But what I’ve realized is that while junior bankers may get some exposure to these situations, far and away their most important responsibility is producing work product. The slide decks and excel models that are created by analysts and iterated on by the team are crucial for articulating the advice that senior bankers provide to their clients.

The rise of large language models (LLMs) like ChatGPT has the potential to revolutionize analyst workflows. Rogo closed its $18.5 million Series A round earlier this month, in a funding round led by Keith Rabois at Khosla Ventures, and are focused on building Wall Street’s first “AI Analyst”. Here is why I have conviction in the business:

1. What is Rogo?

2. The Team Behind Rogo

3. Why is Now Rogo’s Time to Shine?

4. What is Rogo’s Competitive MOAT?

5. Why I Believe in Rogo

1. What is Rogo?

Rogo is building Wall Street’s first “AI analyst” using a robust four-layer technology stack. I’ve tried to visually lay this out below:

At its core, the Data layer forms the backbone of Rogo’s tech stack. The platform is currently powered through two main data sources. The first is proprietary data, acquired through strategic partnerships with financial data providers like S&P Global and Refinitiv. The second source is internal firm data, which Rogo accesses in a highly secure and compliant manner, enabling it to utilize precedent slide decks, excel models, and IC memos within the organization. Soon, Rogo will be able to integrate with the full corpus of an organization’s data, including Slack messages, Outlook emails, and files on internal drives providing Rogo with the same resources as its human analyst counterpart.

The Search layer is designed to organize and make sense of the vast corpus of data it collects. At this layer, Rogo processes and indexes both structured and unstructured data from proprietary financial sources and internal firm databases. This data is then woven into a knowledge graph, where each company acts as a node and is enriched with interconnected data points, such as a product description, financial metrics, and stock price etc. By aggregating data at the company level, Rogo’s retrieval-augmented-generation (RAG) system allows for insights far faster than human analysts. Simply put, Rogo’s search infrastructure can locate a specific data point in a company’s 10-K faster than it would take me to find the same information on Bamsec!

The Chat layer provides a natural, conversational interface that allows users to interact seamlessly with the extensive knowledge graph. This layer functions like an intelligent assistant, enabling professionals to ask questions in everyday language and receive instant, contextually relevant responses. Leveraging the interconnected data from the Search layer, the Chat layer can provide insights at the company level. This allows users to dive into specific queries, such as recent revenue trends or a competitive landscape overview and follow up with additional questions without losing context.

Rogo’s Chat layer outperforms a human analyst due to the iterative nature of the workflows in an investment bank. Often, the initial request from a Partner or Managing Director is not what ultimately makes it into the final set of materials. With each change in question or focus, a human analyst must restart the data collection and search process. In contrast, Rogo’s Chat layer seamlessly handles these shifts in real-time, with its rapid data pipeline, advanced search and indexing capabilities, and contextually relevant interface. This agility allows Rogo to respond instantly to new inquiries, making it significantly more efficient than human analysts.

The agent layer is the pinnacle of Rogo’s tech stack where data insights translate into actionable outputs. This layer autonomously generates polished deliverables such as pitch decks, financial models, and investment memos by harnessing data from the knowledge graph and insights gathered through the chat layer. Unlike traditional workflows, where an analyst would need to manually compile, format, and review information, the agent layer creates complete, presentation-ready materials tailored to industry standards and specific firm templates.

Let’s take an actual example – a common task in investment banking is to benchmark research analysts’ projections for all the key line items for the company of interest. Let’s see how Rogo fairs against its human analyst counterpart.

Let’s first break down the workflow typically required of a human analyst. The process begins with the analyst accessing a database, such as Refinitiv, to download all recent broker reports for a specific company. Next, they must use excel to manually input data for each line item – such as revenue, EBITDA, and net income – that is provided by each broker. This data entry process is repeated for each research analyst covering the company, with most companies often having 15-20+. After compiling this information, the analyst is tasked with transforming this data into a visual format, most likely bar charts, to illustrate the projected metrics from each broker. The final work product would be bar charts for each of the metrics that are benchmarked, which is critical in understanding how each broker views the company.

As a current investment banking analyst who has gone through this process countless times, I can personally attest that it often can consume the better part of a day to gather the data, manually input figures, and create the bar chart outputs. Now, let's explore how Rogo's capabilities can streamline and enhance this workflow.

Instead of manually downloading broker reports and extracting key metrics line by line, Rogo’s full stack solution can handle these steps automatically. Rogo can instantly pull the most recent broker reports from integrated databases like Refinitiv and extract relevant line items directly into a structured format. The platform’s knowledge graph organizes and cross-references each metric by broker, eliminating the need for manual data entry.

Once the data is aggregated, Rogo’s agent layer goes a step further by autonomously creating the final bar chart work product. This entire workflow – from data retrieval to chart creation – is executed within minutes. Not only is this valuable time saved, but Rogo also minimizes errors inherent in manual data handling, providing a reliable, efficient, and error-free alternative to the human analyst.

2. The Team Behind Rogo

Rogo is Wall Street’s next “AI analyst” built by former Wall Street analysts.

Gabriel Stengel (CEO), John Willett (COO), and Tumas Rackaitis (CTO) co-founded Rogo in 2022, driven by their shared vision to transform financial workflows through AI.

Gabriel Stengel holds a bachelor’s degree in computer science from Princeton University. During his time in college, he founded a company focused on automating the creation of pitch decks for investment banks, an early version of what eventually became Rogo. His senior thesis centered on deep learning techniques for interpreting natural language queries in econometrics, highlighting his demonstrated interest in advancing AI applications in finance. Most recently, he worked at Lazard in a cross-functional role, contributing to both the investment banking team and the IT/data team making him uniquely qualified to lead Rogo.

John Willett holds a dual degree in Economics and Computer Science from Princeton University. Before launching Rogo in 2022, he worked as an Investment Banking Analyst at Barclays and J.P. Morgan.

Tumas Rackaitis holds a degree in Computer Science from Oberlin College. Before co-founding Rogo in 2021, Tumas worked as a technology specialist at Gilder Gagnon Howe & Co., where he gained valuable insights into the financial sector and the technological needs of investment professionals. His experience in applying data science to real-world financial challenges enables him to lead Rogo’s development of advanced AI solutions that streamline and enhance financial workflows.

Gabriel and John are finance-oriented founders with a deep understanding of the challenges they aim to address while Tumas brings a strong technical background that positions him well to lead product development. They have been working on Rogo since before the public release of ChatGPT, which has provided them with valuable experience in addressing these challenges from the beginning, prior to the emergence of the current wave of AI advancements. I have strong conviction in this team's capabilities and firmly believe in their visionary goals for the company's future.

3. Why is Now Rogo’s Time to Shine?

There are two reasons why I believe now is Rogo’s time to shine – advancements in the foundational model layer and recent pressure on CTOs to develop their AI strategy.

The earliest versions of Rogo’s product were built prior to the emergence of large foundational models and lacked product market fit​. Initially, Rogo focused on merely providing natural language processing (NLP) for structured and unstructured data sets relevant to finance professionals, which ultimately lacked the novelty needed to attract potential customers​.

Today, Rogo’s product leverages the foundational infrastructure of LLMs, transforming how financial data is processed and interpreted. By fine-tuning these LLMs for financial institutions, Rogo enables its platform to deliver instant answers to complex, vertical-specific workflows. This customization significantly enhances the model's inference capabilities, making it more adept at understanding and generating content that satisfies the needs of finance professionals compared to generic models like ChatGPT.

As foundational models help enable the technological infrastructure necessary for Rogo’s product, CTOs are under increasing pressure to develop their AI strategy.

Since the release of ChatGPT in November of 2022, the number of AI-related sentences in earnings transcripts have spiked dramatically – as shown in the below chart by the Federal Reserve.

Remarkably, in the year after the release of ChatGPT, there was a fivefold increase in sentences mentioning AI per earnings call.

AI is no longer a part of the far-too-distant future; it’s starting to become a tangible topic that is requiring the CTO to act. When it comes to adopting a new AI strategy for a large enterprise, there are three main options the CTO can consider – buy, build, and / or partner.

Buying an off-the-shelf product that can provide immediate value on generic tasks or specific workflows (such as meeting transcribers or ChatGPT) can be an effective strategy for some enterprises, but not financial institutions. These organizations deal with highly sensitive client information that must be handled securely and in compliance with stringent regulations. Any AI strategy implemented by a financial institution must account for the sensitive nature of the data they manage and requires thorough vetting of technology providers by the internal tech team to ensure compliance and security standards are met.

Alternatively, financial institutions can develop internal tools tailored to their own vertical-specific workflows. While this approach may seem effective for large enterprises and bulge bracket investment banks, it overlooks a critical challenge in creating any AI tool: the need for skilled human talent to build it. With tech giants like Microsoft and OpenAI competing fiercely for top AI researchers and possessing the resources to attract them, it’s unlikely that an investment bank or financial institution will have the technical expertise required to develop an AI solution at scale.

I believe most CTOs at financial institutions will focus on the only other option – partnering with a third-party company such as Rogo.

This approach presents an attractive solution by addressing many barriers associated with developing an AI strategy. Third-party providers typically adhere to the highest industry security standards, enabling seamless and compliant data integration. Moreover, because AI is a core component of their offerings, these companies are likely to invest significantly in attracting top AI talent. This allows financial institutions to leverage advanced AI capabilities without the need to hire specialized personnel themselves.

4. What is Rogo’s Competitive MOAT?

CTOs of financial institutions will have three potential partners to choose from – directly leveraging large foundational models that power Rogo’s platform, collaborating with legacy financial data providers, or engaging with startups specializing in AI for financial institutions.

Large foundational models, such as ChatGPT, may struggle to enter this market due to the scaffolding that Rogo provides on top of base LLMs to provide a usable product.

There are three barriers to entry that prevent foundational models from entering Rogo’s markets – data partnerships that Rogo has with financial data providers, best-in-class security protocols for enterprise customers, and stronger knowledge of what financial institutions are looking for in their product. Foundational models are currently trained on large swaths of publicly available data found on the internet. Given that Rogo physically must purchase data from large data providers to fine-tune their models, they offer a superior offering to what an LLM would provide without any customization. In order for LLMs to replace Rogo it would need to do the grunt work of forming the data partnerships that Rogo already has in place.

Even if LLMs are willing to invest the time and resources needed to form these data partnerships, they will need to significantly invest in the compliance and security infrastructure needed to be a technology provider to financial institutions.

Rogo offers a product that complies with both GDPR and FINRA regulations, ensuring it meets the highest security standards in the industry. Built on a modern zero-trust security model, Rogo incorporates principles such as least privilege access, just-in-time access management, and robust authentication procedures. All data handled by Rogo is encrypted in transit using TLS 1.2, while data at rest is secured with AES-256 encryption. Additionally, Rogo maintains a dedicated on-call security team that provides 24/7/365 coverage to address any potential security incidents, ensuring the integrity and confidentiality of sensitive financial information

LLMs would need to build or integrate a similar security infrastructure to compete with Rogo’s offering.

Lastly, LLMs lack the deep understanding of the specific enterprise use case that Rogo is looking to address. The founders of Rogo bring firsthand experience from their time in investment banking, having directly engaged with the challenges of the analyst tech stack. This competitive advantage allows Rogo not only to develop the most relevant and impactful AI agents but also to foster relationships with key decision-makers at investment banks. These connections are crucial for facilitating the implementation of Rogo’s product throughout an organization and ensuring that it meets the unique needs of financial professionals.

While foundational models at least have the technological capability of replacing Rogo, legacy players are still far behind.

Given the heightened sensitivity of client data and the fact that financial institutions are primarily focused on their core business rather than technology, many have struggled to keep pace with the latest advancements in their tech stacks. Currently, platforms like Capital IQ and FactSet are among the most relied-upon financial information databases, providing essential data for investment banking analysis. However, both platforms are notoriously challenging to navigate, often requiring complex excel formulas with intricate syntax to extract even a single data point. Additionally, these platforms are not always accurate, leading analysts to frequently revert to primary sources to verify each number before including it in their slide decks and excel models.

I believe the relatively poor quality of the current tech stack at financial institutions will drive the adoption of Rogo. Rogo’s product provides a fast and efficient way to trace the source of any data point it delivers, similar to the functionality of Perplexity. This feature allows users to quickly audit the information provided, significantly reducing the time spent on verification. Even if Rogo's outputs are incorrect – much like the inaccuracies often found in Capital IQ and FactSet – analysts can swiftly cross-reference and validate numbers, enhancing their overall efficiency and reliability in data analysis.

While there are a handful of other AI startups in the financial institution space, none of them offer a similar product to Rogo.

Ultimately, Rogo’s true differentiator lies in the fact that it was built for investment banks by former investment bankers. While Hebbia is a notable player at the intersection of AI and financial institutions, having raised a $130 million Series B in July 2024, its product is not specifically tailored to the needs of investment banking and financial institutions. Rogo’s core offer includes its PowerPoint agent, which addresses a critical aspect of the investment banking workflow and where bankers spend most of their time. To my knowledge, no competitors, including Hebbia, offer a similar solution that is so finely tuned to the specific demands of investment banking professionals, making Rogo the go-to choice in this market.

5. Why I Believe in Rogo

As a current investment banking analyst, I’ve witnessed firsthand the challenges of producing critical work product with an outdated technology stack. While the rise of LLMs presents an opportunity to revolutionize analyst workflows, Rogo stands out as the first “AI Analyst” specifically designed for the unique demands of financial institutions. With a robust four-layer technology stack, Rogo efficiently procures, stores queries and interprets data, significantly reducing the time and effort required for analysts to create work product.

According to McKinsey, global technology spending by financial institutions reached $650 billion in 2023 and is projected to grow at an annual rate of 9%. Assuming Rogo is able to capture 0.5% of this market, their Series A post money valuation of $80 million represents a > 40x opportunity for investors. Rogo has already gained significant traction, with 25 financial institutions currently using its product. In various interviews, the founders have indicated a strong pipeline of potential customers, acknowledging that integration takes time due to the necessity for compliance and secure handling of sensitive firm data.

The primary existential risk for Rogo lies in the mishandling of sensitive firm data. While Rogo offers a highly secure solution, a single security breach could jeopardize the entire company. To gain full conviction in this business, I would ideally want to speak with financial institutions directly to understand their perspectives on data risk when using a third-party vendor like Rogo.

Thanks for making it all the way to the end! If you have any thoughts, questions, or feedback, I'd love to hear them – your input is always valuable.